(a) Overview--The latest large-format printer-plotters, and smaller desktop printers as well, have a very strong need for throughput. The principal current objectives include:
making printheads as long as possible, in the printing-medium advance direction ("PAD"), to increase the dimension of each swath in that same direction; and PA1 reducing the number of passes. PA1 first group of M bands, odd and even, PA1 second group of M bands, odd and even, PA1 next group of M bands, odd and even, PA1 . . . PA1 next group of M bands, odd and even, PA1 (N-1)th group of M bands, odd and even, PA1 Nth group of M bands, odd and even. PA1 the first group, odd, and the second, fourth, sixth and eighth groups, even; PA1 the first group, even, and the second, fourth, sixth and eighths groups, odd; PA1 the second group, odd, and the third, fifth and seventh groups, even; PA1 the second group, even, and the third, fifth and seventh groups, odd; PA1 the third group, odd, and the fourth, sixth and eighth groups, even; PA1 the third group, even, and the fourth, sixth and eighth groups, odd; PA1 the fourth group, odd, and the fifth and seventh groups, even; PA1 the fourth group, even, and the fifth and seventh groups, odd; PA1 the fifth group, odd, and the sixth and eighth groups, even; PA1 the fifth group, even, and the sixth and eighth groups, odd; PA1 the sixth group, odd, and the seventh group, even; PA1 the sixth group, even, and the seventh group, odd; PA1 the seventh group, odd, and the eighth group, even; PA1 the seventh group, even, and the eighth group, odd.
These objectives are intimately related, since the first is a primary means of accomplishing the second.
Longer heads alone, however, are not the end of the matter. Even introducing a printhead long enough to physically print with a minimum number of passes would leave many problems of printing quality unresolved.
Both the above objectives do penalize image quality, for at least three reasons. First, longer printhead is more difficult to manufacture, and has more chances for lower performance than the average in some localized area along the head. Such conditions may occur in the head when it is first manufactured, or may arise later during the life of the head.
Second, a smaller number of passes necessarily requires application of a greater ink density in each pass. Therefore ink density has a larger discontinuity at swath boundaries, and coalescence at the boundaries accordingly worsens.
Third, a smaller number of passes, in conjunction with a longer swath in the PAD direction, produces banding at lower spatial frequencies. Band patterns at these spatial frequencies is more unpleasant to the human eye.
Some recent and concurrent efforts attempt to address the swath-boundary coalescence problem. For instance, different ink-density profiles have been introduced at or near swath boundaries.
These efforts have improved performance; nevertheless, each such effort has addressed primarily just one of the three penalties just noted. Integration of solutions to all three is an objective of the present document. In addition, Joan Manel Garcia has introduced methods for generation of masks that assign different workloads to each nozzle--and this document bears on the selection of input data for use in his mask-generation method.
(b) Spatial-frequency effects in banding--A persistent problem in incremental printing is conspicuously visible banding or patterning, which arises from a great variety of causes. Generally these causes are associated with repetitive phenomena that are inherent in the swath-based natured of such printing.
Garcia, in U.S. utility-patent applications Ser. Nos. 09/150,321 through '323, particularly addresses problems of patterning in the lateral or transverse dimension, i. e. parallel to the scan axis. He points out that such patterning is especially objectionable when it occurs at spatial periodicities to which the human eye is particularly sensitive.
Garcia shows that such banding can be rendered very inconspicuous at normal reading distances by moving its periodicity to roughly 3 cm (1 inch), or preferably a bit longer. This can be accomplished by tiling printmasks of those widths.
Unfortunately that technique is not now readily applicable to the longitudinal dimension--i. e. to the direction parallel to the print-medium advance axis. The reason is that, generally, largest current-day printheads are only about 21/2 cm (1 inch) long in that direction.
Within the corresponding available range of spatial frequencies, banding in the lower three-quarters of that range (used in single-pass through four-pass printmodes) is quite conspicuous. Unfortunately the current trend toward reducing the number of passes used for printing each image segment--to enhance overall printing throughput --militates toward use of precisely that part of the range.
(c) Swath-interface effects--Some banding along the print-medium advance axis arises at the interfaces between swaths--due to the advance errors and "PAD" errors mentioned above, and due to ink-media interactions such as coalescence or print-medium expansion. Earlier documents such as Doval's have pointed out that repetitive, small failures of abutment themselves introduce banding (though extremely tiny imprecisions or variations in abutment can be helpful).
Swath-abutment irregularities may represent the single most conspicuous form or type of banding effect. When one swath edge is closely abutted to another, the abutment is almost always imperfect--leading to either a shallow gap between swaths or a shallow overprint where they overlap.
Also the two swaths are generally not exactly the same in darkness or color saturation, adding another element of contrast along the interface. Such problems are aggravated by a high or abrupt gradient of wetness along the edge of a just-deposited swath, when an abutting swath is formed soon after.
So-called "PAD error" has attracted particular attention in part because some modern pens are subject to a concentration of aiming errors at the ends of the pen--most classically outboard-aimed nozzles 91 (right-hand "A" view, FIG. 7) as distinguished from the great majority of more centrally disposed nozzles 90.
This higher density of errors, with systematic outboard aim, results from the greater difficulty of maintaining TAB-tape nozzle arrays planar, in comparison with the metal nozzle plates used earlier. In some heads, particularly at the ends of the array, the tape is typically wrapped around the adjacent ends of the printhead--causing the tape to curl very slightly.
The outboard aim in pens of this type increases 93 the overall dimension of the pixel swath in the print-medium-advance axis, beyond the nominal width 92. Typically this overall increase has been on the order of two or three rows.
As a result, when adjacent swaths that should neatly abut are printed with a nominal advance of the print-medium-advance mechanism, those swaths will instead overlap slightly. This occurs because an error region 93 in one of the swaths projects into a region which should be occupied by the other swath.
(d) Internal effects--Not all banding problems, however, occur at swath boundaries. Some result simply from nozzle PAD problems, and these nozzle irregularities can be entirely internal to the swath (right-hand "B" view, FIG. 7).
Internal patterns, in turn, can be formed by repetitive coincidences of nozzle irregularities. Prior systematic procedures placed particular irregularly-performing pairs (or other groups) of printhead elements into conjunction--with respect to the printing medium--over and over.
As an example, the Hewlett Packard Company printer product known as the Model 2000C uses two-pass bidirectional printmodes--each pixel row being printed by two separate nozzles. At 24 rows per millimeter (600 dots per inch, dpi), a 12.7 mm (half inch) pen, has 300 nozzles.
Ordinarily nozzles number 1 and 151 contribute drops to the same image row--using a 61/3 mm (quarter inch) advance and, again, a two-pass, 300-nozzle printmode. Every 61/3 mm these same two nozzles are paired (see the above-mentioned Zapata patent document, particularly in that document FIG. 7 and the Table).
If nozzles 1 and 151 when used in combination form a noticeable band effect, this effect is highly visible to the user--because it is present in a repeating pattern, roughly every 6 mm or quarter inch. For example, if both nozzles happen to be directed well away from their nominal target pixel row, then that pixel row will appear unprinted (at least in the particular color in which the head in question prints), rather than the nominal double-printed.
Another kind of band effect can be caused by an interaction of nozzles that are adjacent or nearby. For example assume that nozzle number 5 is aimed "low" (toward the nominal target row for nozzle 6). If nozzle 6 is aimed accurately, its target row will be double-printed.
If in addition nozzle 156 is also aimed accurately but nozzle 157 is aimed "high" (i. e. both toward the target row for nozzle 156), then in the printed image the common pixel row for nozzles 6 and 156 will be quadruple-printed --while the adjacent rows above and below will each be single-printed rather than the nominal (double printed).
In short, banding within swaths results from repetitive coincidences between irregularly printing elements within each combination. Patterning arises from repetitive, systematic operation.
Objectionable patterning is subject to quantitative effects. Thus some printmasking approaches to patterning in effect simply dilute repetition within an environment of a greater number of alternative states.
(e) Multipass printmode solutions--Heretofore a common strategy for dealing with all these problems has been to increase the number of passes used to print each image segment. This strategy, however, degrades printing throughput.
It is therefore disadvantageous in the present market, which is increasingly more demanding. This marketplace is characterized by continuously escalating consumer perceptions of what constitutes an acceptable overall image-printing time.
(f) Conclusion--Thus failure to effectively address problems of banding in printmodes using low numbers of passes has continued to impede achievement of uniformly excellent inkjet printing--at high throughput. Thus important aspects of the technology used in the field of the invention remain amenable to useful refinement.